Ben Woodard <ben_at_valinux.com> wrote:
> 
> Jan Harkes wrote:
> > Check if there is anything mapped in memory between 0x20000000 and
> > 0x2c8000000. On linux when the server is hanging in it's crashed state,
> > do "cat /proc/`pidof codasrv`/maps".
> > 
> 
> Hmm. that appears to be the problem. All the shared libraries are
> loaded there. Why do you think that is? The only thing that I can
> think of is this. On VA machines, the default kernel is patched so
> that you can have more memory in a machine >1Gb. However, this cuts
> down on the address space available for a process. It looks like I am
> running one of those custom kernels. However, I would almost swear
> that I was running one of those for as long as I have been running coda.
> 
> [ben_at_wythe ben]$ sudo cat /proc/1070/maps
> 08048000-08129000 r-xp 00000000 08:04 143495     /usr/sbin/codasrv
> 08129000-0813a000 rw-p 000e0000 08:04 143495     /usr/sbin/codasrv
> 0813a000-08165000 rwxp 00000000 00:00 0
> 15000000-15004000 rw-p 00000000 00:00 0
> 15006000-1500a000 rw-p 00000000 00:00 0
> 2aaab000-2aabd000 r-xp 00000000 08:04 30722      /lib/ld-2.1.1.so
> 2aabd000-2aabe000 rw-p 00011000 08:04 30722      /lib/ld-2.1.1.so
> 2aabe000-2aabf000 rwxp 00000000 00:00 0
> 2aabf000-2aac0000 r-xp 00000000 08:04 30726      /lib/libNoVersion-2.1.1.so
> 2aac0000-2aac1000 rw-p 00000000 08:04 30726      /lib/libNoVersion-2.1.1.so
> 2aac1000-2aac4000 rw-p 00000000 00:00 0
> 2aac6000-2aad2000 r-xp 00000000 08:04 30736      /lib/libdb1-2.1.1.so
> 2aad2000-2aad4000 rw-p 0000b000 08:04 30736      /lib/libdb1-2.1.1.so
> 2aad4000-2ab05000 r-xp 00000000 08:04 74099      /usr/lib/libstdc++.so.2.8.0
> 2ab05000-2ab13000 rw-p 00030000 08:04 74099      /usr/lib/libstdc++.so.2.8.0
> 2ab13000-2ab15000 rw-p 00000000 00:00 0
> 2ab15000-2ab30000 r-xp 00000000 08:04 30740      /lib/libm-2.1.1.so
> 2ab30000-2ab31000 rw-p 0001a000 08:04 30740      /lib/libm-2.1.1.so
> 2ab31000-2ab32000 rw-p 00000000 00:00 0
> 2ab32000-2ac18000 r-xp 00000000 08:04 30729      /lib/libc-2.1.1.so
> 2ac18000-2ac1d000 rw-p 000e5000 08:04 30729      /lib/libc-2.1.1.so
> 2ac1d000-2ac20000 rw-p 00000000 00:00 0
> 2ac25000-2ac2d000 r-xp 00000000 08:04 30760      /lib/libnss_files-2.1.1.so
> 2ac2d000-2ac2e000 rw-p 00007000 08:04 30760      /lib/libnss_files-2.1.1.so
> 2ac2e000-35cce000 rw-p 00000000 00:00 0
> 7fffd000-80000000 rwxp ffffe000 00:00 0
> 

Yes, I think this is the problem.  You can pick a different "rvm starting
address" -- instead of using the default 0x20000000 (for linux) , use
something like 0x50000000 (or other) with rdsinit.  You have two methods
to use rdsinit and rvmutl: 1. run them manually (then read rdsinit(8)
and rvmutl(8) for more details); or 2. use a modified version of
vice-setup-rvm.  Of course, the script vice-setup-rvm can serve as an
example if you want to run the two programs manually.

Good luck!

-- Clement